150 likes | 412 Views
The Garden Sprinkler Effect. Hendrik Tolman. The WAVEWATCH III Team + friends Marine Modeling and Analysis Branch NOAA / NWS / NCEP / EMC NCEP.list.WAVEWATCH@NOAA.gov NCEP.list.waves@NOAA.gov. Outline. Covered in this lecture : What’s the Garden Sprinkler Effect (GSE)?
E N D
The Garden Sprinkler Effect Hendrik Tolman The WAVEWATCH III Team + friends Marine Modeling and Analysis Branch NOAA / NWS / NCEP / EMC NCEP.list.WAVEWATCH@NOAA.gov NCEP.list.waves@NOAA.gov
Outline • Covered in this lecture: • What’s the Garden Sprinkler Effect (GSE)? • Present solutions. • Hurricane Florence example. • Multi-grid hurricane example. • Notes and future.
GSE • Definition: • Incompatibility in various discrete resolutions in model. • Spectral resolution in direction is so coarse, that spatial propagation of discrete directions results in discrete disintegration of swell field that should be continuous, Booijand Holthuijsen (1987, JCP).
GSE Swell propagation from lower left corner over 5 days (ww3_tp2.3) Exact solution: continuous dispersion of swell energy over a large area. (no dissipation !)
Third order accurate Ultimate-Quickest scheme (Leonard) of WAVEWATCH III. Obvious garden sprinkler effect, spectral discretization results in disintegration of swell field. Better scheme gives worse results. Essentially uselessin this form. GSE Swell propagation from lower left corner over 5 days (ww3_tp2.3)
Solutions • Traditional: • Old method (Booij and Holthuijsen, 1987). • Add diffusion term to represent sub-grid dispersion. • Good but excessively expensive for small scale models (NAH). Requires filtering at high latitudes. • Newer: • Replace diffusion with straightforward averaging. • Almost same as original. • Much cheaper than old method.
GSE Swell propagation from lower left corner over 5 days (ww3_tp2.3) Exact solution: continuous dispersion of swell energy over a large area.
Third order accurate Ultimate-Quickest scheme (Leonard) of WAVEWATCH III. Obvious garden sprinkler effect, spectral discretization results in disintegration of swell field. GSE Swell propagation from lower left corner over 5 days (ww3_tp2.3)
UQ scheme with Booij and Holthuijsen (1987) diffusive dispersion correction. Major improvement over plain UQ scheme, tunable. Due to explicit schemes, stability becomes a major issue at small grid steps (order 25 km). GSE Swell propagation from lower left corner over 5 days (ww3_tp2.3)
UQ scheme with simple pre- or post- averaging of fields. Virtually identical results as previous, tunable, cheap. "exact" dispersion area discrete advection linearized dispersion area GSE Swell propagation from lower left corner over 5 days (ww3_tp2.3) Tolman, 2002: Ocean Mod., 4, 269-289.
Hurricane Florence UQ plain 1.00 UQ + dif. 1.75 UQ + avg. 1.09 Peak periods from 7 to 10s from hurricane Florence at 00z Sept. 13 2000 from NCEP's NAH model. Relative computational costs in red.
insufficient resolution borderline resolution sufficient resolution inconsistent results Garden Sprinkler Effect best solution Multi-grid hurricane Stationary hurricane with default settings in WAVEWATCH III. 50km grid 15km grid 5km grid composite of grids large 5km grid multi-grid model
Notes and future • There is some GSE mitigation, but no solution. • As the resolution of models goes up, the GSE shows up again. • Example of multi-grid model and automatic scale adjustments that helps. • Not easily applied generally. • Alternative method to solve uses divergent dispersion consistent with mean energy per spectral grid box. • Experimental in Tolman (2002). • Needs to be set up using local spectral data only to become feasible. • Is this a solution or a mitigation?
The end End of lecture